Thread by @gshotwell

This is an open label study of 584 Brazillian Covid patients. Study participants took either hydroxychloroquine, nitazoxanide or ivermectin and none (!!) of them went to the hospital or died.

researchsquare.com/article/rs-981…

Now there's certainly something wrong with this study. 16% of Brazillian Covid patients go to the hospital so the odds of finding one drug that eliminates hospitalization are small; the odds of finding three in one trial are basically zero.

The reason I'm tweeting about this is that 80% of the patients in the trial were given vitamin D according to clinical abnormalities. I wrote to the author to ask what those were and what dose was given. Could vitamin D confound a study this badly?

Our prior for vitamin D and Covid should be that it helps a little bit with respiratory infections, but only using long term, low dose regimens.

bmj.com/content/356/bm…

One reason to have this modest prior is, as @methodsmanmd writes, vitamin D almost always flames out when you get higher quality data.

medscape.com/viewarticle/93…

The odd thing about the Covid research is that the higher quality studies have pointed to very strong effects using short term therapy. The big observational trials and intervention papers have all found much larger effect sizes than we'd expect given the prior.

For example this one showed that vitamin D supplementation eliminated ICU visits among elderly hospitalized patients.
sciencedirect.com/science/articl…

I still think it's pretty unlikely that vitamin D will act this strongly on Covid, but the data we have so far, flawed as it is, says that an effect that big is plausible.

This means that, if you're running a Covid trial, you should be careful about which patients are taking vitamin D. Giving it to everyone in the trial might totally screw up the inferences you can draw from the data.

In this case it's possible that all three of these drugs act strongly alone, or that they act in concert with vitamin D, or that they do nothing and vitamin D is the main cause. In other words, we don't really learn that much from it at all.